Stereo matching with temporal consistency using an upright pinhole model

Jiang, Yibo

Permalink

https://hdl.handle.net/2142/97862 Copy

Description

Title

Stereo matching with temporal consistency using an upright pinhole model

Author(s)

Jiang, Yibo

Contributor(s)

Do, Ming N.

Issue Date

2017-05

Keyword(s)

• stereo vision
• temporal consistency

Abstract

Stereo vision, as a subfield of computer vision, has been researched for over 20 years. However, most research efforts have been devoted to single-frame estimation. With the rising interest in autonomous vehicles, more attention should be paid to temporal consistency within stereo matching as depth matching in this case will be used in a video context. In this thesis, temporal consistency in stereo vision will be studied in an effort to reduce time or increase accuracy by utilizing a simple upright camera model. The camera model is used for disparity prediction, which also serves as initialization for different stereo matching frameworks such as local methods and belief propagation. In particular, this thesis proposes a new algorithm based on this model and sped-up patchMatch belief propagation (SPM-BF). The results have demonstrated that the proposed method can reduce computation and convergence time.

Type of Resource

text

Language

en

Permalink

http://hdl.handle.net/2142/97862

Owning Collections